Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Tax Policy and Economic Growth in Jamaica Hubert G. Scarlett1 Research Services Department Research & Economic Programming Division Bank of Jamaica Abstract Numerous studies have indicated that tax policies aimed at increasing government revenue have being regarded as an impediment to economic growth in both the short and long run. The research at hand, seeks to explore the impact that increase taxation from any tax measure may have on economic growth in Jamaica. This objective is achieved by using a general autoregressive distributed-lag model, which jointly captures both short and long run effects. Unlike other studies that utilized annual data, the impact will be captured using quarterly data. Additionally, an attempt to ascertain the directional relationship between the explanatory variables and economic growth is done using the Granger causality test. The findings indicate that increasing revenue inflow from indirect taxes prove to be more conducive to economic growth in the long run. On the other hand, increasing the share of taxes from personal income (P.A.Y.E.) has the greatest harm on per capita GDP over time, and correction to equilibrium from such an impact would take up to nine months. 1 The views expressed in this paper are those of the author and in no way represent an official position of the Bank of Jamaica. Table of Contents 1.0 Introduction .......................................................................................................................... 3 2.0 Theoretical Perspective of Tax Policy and Economic Growth ............................................ 4 3.0 Literature Review................................................................................................................. 5 4.0 Data & Methodology ........................................................................................................... 7 5.0 4.1 Data ............................................................................................................................... 7 4.2 Description of the Model ............................................................................................ 10 Results and Discussions ..................................................................................................... 12 5.1 Growth Components: Baseline Equation ................................................................... 13 5.2 Growth Model with Tax Measures ............................................................................. 14 5.3 Growth Model with Taxation Index ........................................................................... 16 5.4 Discussion ................................................................................................................... 17 6.0 Conclusion & Recommendations ...................................................................................... 18 7.0 References .......................................................................................................................... 19 8.0 Appendix ............................................................................................................................ 20 2 1.0 Introduction Between 2008 and 2010, the Jamaican economy contracted on average by 1.7 per cent per annum, following average annual growth of 2.1 per cent over the previous five years. A combination of factors contributed to the poor performance of the economy over the last three years, chief among them being the global financial crisis. At present, one of the primary objectives of the authorities is to restore growth to the Jamaica economy, both in the short, medium and long-term. In this context, the aim of this paper is to analyse the effectiveness of different tax policy measures given the current economic crisis as well as the ability of the measures to contribute to long run growth in the Jamaican economy. The findings from this study are expected to play an advisory role to the Government in its formulation of tax policy. Tax policies from time to time have been implemented for a variety of reasons. According to Romer and Romer (2010) tax policies are implemented either to: (i) finance a budget deficit, (ii) promote long run growth, (iii) pay for increased government spending and (iv) counter other influences in the economy. In the case of Jamaica, and that of many other developing countries, with relatively high public debt to gross domestic product (GDP) ratio, tax policies have for the most part been geared towards filling a financing gap to supplement borrowings. The tax measures implemented had varied impacts on the economy, some more harmful than others. Kneller, Bleaney and Gemmell (1999) suggest that in designing tax policies to promote growth, greater reliance should be placed on indirect taxation. This is seen as being less harmful to the economy as it does not reduce the return on investment when compared to direct taxation, which presents a disincentive to invest in physical and human capital. Over the last decade, Jamaica’s tax composition favours indirect taxes, which on average accounted for approximately 52.2 per cent of tax revenue per annum. This bias is essentially in line with the taxation characteristic for many developing countries. The paper uses a standard growth function, within an autoregressive distributed lag framework, with quarterly data from 1990 to 2010, to assess the impact of taxation on economic growth. The research does not discount the fact that tax measures could be a function of the economic condition, a very familiar situation for Jamaica and other small open economies. . 3 The remainder of the paper is structured as follows. Section two presents the theoretical perspective of tax policy and economic growth, while section 3 presents the literature review. Section four presents the data incorporated in the study along with the framework employed to assess the impact of tax policy on per capita GDP. This is followed by section five that discusses the empirical findings while section six gives the closing remarks and the policy recommendations. 2.0 Theoretical Perspective of Tax Policy and Economic Growth Fiscal policy, on a whole, is geared towards having an impact on demand in the long run, with monetary policy assisting in the short run by maintaining price stability. Within the context of tax policy, tax incentives and holidays are offered to sectors of the economy in an effort to spur growth or to protect the small and vulnerable enterprises from outside competition. Whatever the tax measure employed, it has to be complemented by other fiscal and developmental measures that are in line with the long-term objective of the economy. It has been argued by the Keynesians that reducing the tax rate on direct taxation, in particular personal income will transfer greater spending power to the individual, thus facilitating an increase in consumption expenditure and hence economic activity. This increases savings and or investment opportunities. On the other hand, any attempt to increase tax collected from direct taxation may serve as a disincentive to work, that is a reduction in the supply of labour, and a switch to greater leisure and hence a reduction in productivity. The Ricardians, however, argues that if taxes are reduced, it will require the Government to balance its budget in the short run by increasing borrowing. These views, however, imply that the tax policy is not revenue neutral. Arnold et al (2011) posits that tax changes that are geared towards innovation and entrepreneurship may have persistent and positive long run growth effects. However, the implementation of any tax policy has to bear in mind the potential fiscal imbalance of the country. Additionally, the impact that a tax policy is likely to have on economic growth will vary and depend greatly on the country’s development progress and the level of production. As indicated by Jones (2001), the distribution of income towards the factors of production, such as 4 capital and labour, plays a vital role in the level of production. The author also indicated that external factors do contribute to the level of growth in output, given the fact that there exists a close relationship between the growth in output and growth in the volume of international trade. In examining the impact of tax policy on economic growth, the theorists have utilized growth models coupled with varied tax measures to explain the nexus. The neoclassical model of Solow (1956) suggests that tax policy has no impact on economic growth in the long run. This model assumes that key factors of production such as population growth and technological advancement are determined outside the model. On the other hand, endogenous growth theorists, who believe that economic expansion is determined within the system, posit that tax policy do have an impact on economic growth over time. Assuming that output grows over time, Romer and Romer (2010) indicate that government induced tax actions can be taken to offset developments that may cause output to deviate from its normal path. For example, a tax cut may be implemented if policy makers need to stimulate demand to bring the economy out of a recession. Further, tax measures could be implemented to increase growth above its long-term trend. When policymakers consider a tax policy, they have to deliberate on the measures that will have the least negative impact on the economy. For a developing country whose tax structure is bias towards indirect taxes, policymakers have to consider the impact raising the tax rate is likely to have on the general price level. If policymakers consider increasing the share of tax revenue intake from direct taxes, consideration will have to be given to its impact on capital investment. 3.0 Literature Review Tax policies may impact economic activities differently across countries, in both the short and long run. However, the underlying theory suggests that tax policies should have a negative impact on economic growth. Empirical studies have utilized growth models with varied specifications to test the theory to ascertain the directional and degree of impact of tax policies across countries and territories. It has been noted that these policies may have negligible to 5 recognizable impacts. The models often start with the standard growth variables, physical capital, human capital and growth of the labour force to which tax indicators are incorporated. Kneller et al (1999) examined the impact of fiscal policy on growth using an annual panel data set of twenty-two (22) countries within the Organization for Economic Co-operation and Development (OECD) over the period 1970-95. The approach included the complete specification of the government budget constraint (revenue and expenditure), relative to other endogenous growth models, which only incorporate the revenue aspect. This approach as indicated by Kneller et al (1999) built on the methodology proposed by Barro (1990) where the complete specification of the government’s budget constraint corrected for the biases that existed with a partial specification. The model involved regressing of non-fiscal and fiscal variables on economic growth rates, with one component of the fiscal variable being omitted.2 The findings indicate that distortionary taxation (direct taxes) reduced growth while non-distortionary taxes did not. Greenidge and Drakes (2009) utilized an unrestricted error correction model to examine tax policy and its effect on macroeconomic activity in Barbados. The approach initially proposed by Pesaran and Shin (1997) jointly captured both short and long run effects from a general autoregressive distributed-lag model. This procedure, when compared to others, has the ability to assess the cointegration of variables despite their order of integration and better at handling small samples and dynamic source of biases. The model is first estimated with standard growth variables, to which the tax indicators were individually added. The tax indicators were constructed using a tax index and principal component analysis.3 Total and indirect taxation had a contractionary impact on the economy in the short run with no long run impact, while direct taxation had a negative impact on growth in both short and long run. Romer and Romer (2010) investigated the impact of tax changes on economic growth in the United States during the post-war period. Information on the legislative tax changes were gathered from narrative sources (including presidential speeches) and regressed on changes in real GDP over the period 1947 to 2007. The tax changes were separated into those related to 2 The omitted variable is regarded as the implicit financing element and as such the impact is revenue neutral. See Greenidge and Drakes (2009) for further details. The principal component analysis aggregates the individual tax index. 3 6 prospective economic conditions and exogenous (other) reasons.4 The results indicated that tax changes have very large effects on output resulting in reduction of between 2.5 and 3.0 per cent. In addition, output effects were found to be more closely linked to changes in actual taxes rather than to news about future changes. Arnold et al (2011) examined the long run relationship between tax structures and economic growth within the OECD. The authors utilized an error correction model with annual panel data of 51 countries. The model included individual tax indicators (expressed as a share of total tax revenue) along with the typical growth variables (physical & human capital and population growth) as well as a control variable, tax revenue to nominal GDP.5 The explanatory variables, including the lagged dependent variable, were used in both levels and first differences to account for transitional dynamics. The findings indicated that long run economic growth could be improved by gradually increasing taxes on consumption and immovable property as well as improving the design of individual taxes. Personal & corporate income taxes, consumption and immovable property taxes had the least harmful impact on long run GDP per capita. It is also suggested that reducing income taxes on low incomes would be the best option for increasing economic growth and aiding economic recovery. This is in a context where reduced income taxes on low incomes would stimulate demand, increase work incentives and lessen income inequality. 4.0 Data & Methodology 4.1 Data Using a standard growth model framework, the impact that changes in Jamaica’s tax structure have on economic growth is assessed with the use of quarterly data from 1990 to 2010. The data set include GDP per capita, physical capital and human capital as well as growth rate of the labour force and nineteen (19) tax indicators. The measurement for each variable is outlined in Table I in the Appendix. 4 See Romer and Romer (2010) for further details. The control variable eliminates any bias that could result from a correlation between the tax mix and overall tax burden, given the varied tax structure across the countries examined. 5 7 Physical capital, human capital and population growth rate have been regarded as the fundamental factors of economic growth. As indicated by Földvári1 and Bas van Leeuwen (2009) various studies on economic growth have utilized either literacy rates, primary school enrolment, age-heaping or average years of education (most commonly used) as a proxy for human capital. This study uses primary institution enrollment as the proxy for human capital. Literacy rate is not used as a proxy for human capital as it is close to unity and as such would pose empirical problems and lead to inaccurate estimation as indicated by Benhabib and Spiegel (1994). The proxy use for physical capital accumulation is gross fixed capital as a share of GDP, which is widely accepted and also provides information as to the level of reinvestment into the economy and thus its contribution to economic growth. To examine the impact of Government’s tax policy measures, the tax policy variable is defined as the share of tax revenue that is raised from a given tax as proposed by Arnold et al (2011). The tax measure essentially provides an indication of the level of taxation and the policy action of the fiscal authority. However, one has to bear in mind that the tax structure will take into consideration the tax group that policymakers want to place their focus. In this regard, the tax structure is grouped into six groups. Each tax group indicate the change that will result in per capita GDP from an increase in its share of taxes. It is important to note that an increase in the share of tax revenue for one tax group will automatically reduce the amount of taxes required to be raised from the other tax groups. This implies that the tax impact would be revenue neutral. The groups are defined as follows, (the components of each are outlined in Table II in the Appendix) Direct, Indirect, Consumption, International Trade, Income & Profit and Production & Consumption. Figures 1 and 2 below depict the structure of Jamaica’s tax system and the share of each tax group to tax revenue, respectively. 8 Figure 1 Figure 2 Following Greenidge and Drakes (2009), tax indices were created for individual tax rates as another approach to measure tax policy, Indices were only created for the rates of pay as you earn (P.A.Y.E.), general consumption tax (G.C.T.) and corporate income taxes, given the ease of accessing and coding these rates. With the data set starting in the March 1990 quarter a value of 1 is assigned to the starting period to represent the base period. If the tax rate did not exist at that time a value of 0 is assigned until the tax came into effect. Thereafter, as the tax rate changes the percentage change in the burden is estimated and the index adjusted by this changes, thus creating an index series for each tax types. The series created are shown in Figure 3 below.6 Figure 3 6 Tax Index = 1 + percentage change in the tax rate. The index for corporate tax would be unity over the assessment period, as the rate has remained at 33 1/3 %. 9 A descriptive statistics of the variables employed in this study along with their correlation with per capita GDP and order of integration are depicted in Table III in the Appendix. The order of integration is determined using the Phillips-Perron test at 5.0 per cent level of significance.7 In addition, Granger Causality test are carried out to examine the causal relationship between per capita GDP and the explanatory variables. The null hypotheses tested are: (i) economic growth does not Granger cause the ‘explanatory variables’ and (ii) ‘explanatory variables does not Granger cause economic growth. As a testing criterion, the probability value used is at the 5.0 per cent level of significance.8 This analysis will aid in providing some insight into the relationship between taxation and economic growth. 4.2 Description of the Model The model used in the paper is based on a Cobb-Douglas type function, in which output (Y) is a function of physical capital (K), human capital (H), level of technological and economic efficiency (A) and labour (L), as indicate below: Y (t ) K (t ) H (t ) ( A(t ) L(t ))1 …………….[1] where, & are partial elasticity relative to the variable. Variable, A, is dependent on policy actions After log transformation equation 1 models GDP per capita as a function of investment rate, per capita human capital and growth rate of the working population. In an attempt to assess the impact of changes in tax measures on economic growth within the growth model framework, a general autoregressive distributed-lag (ARDL) model is used. Greenidge and Drakes (2009) utilised the ARDL model, developed by Pesaran and Shin (1997) in their assessment of the impact of changes in tax rates on growth in Barbados. The study for Jamaica opted to utilize the ARDL approached instead of an error correction model as utilized by Arnold et al (2011) and others for the following reasons.9 Firstly, the ARDL model is selected on the basis that the model specification has no bias to the order of integration of the variables when assessing a long run 7 Phillip-Perron is said to be best suited for small sample size. Granger causality measures precedence and information. See Eviews User guide for additional explanation. 9 It should be noted that both procedures are capable of accounting for off-equilibrium dynamics of GDP per capita explicitly without losing any information. 8 10 relationship compared to other error correction models, such as the vector error correction model, which requires all variables to be integrated of the same order.10 Secondly, the model has proven to work well with small sample size as it provides statistically significant coefficients, relative to the other measures which are best suited for large samples. Thirdly, the estimation of the model can be carried out using ordinary least squares (OLS) from which the cointegrating relationships can be ascertained. The ARDL framework implemented in its logged format is outlined below: 4 ln y t c0 ln y t 1 a1 ln k t 1 a 2 ht 1 a3 wkpt 1 iVt 1 bi ln y t i i 1 51 51 51 51 i 0 i 0 i 0 i 0 1i ln k t i 2i ln ht i 1i ln wkpt i i ln Vt i t ........... [2] Where y represents GDP per capita, k for physical capita accumulation, h for human capital, wkp for growth rate of the working population, V for policy variables/tax indicators and t for the error term. The coefficients on the level effects are , ai , i with the long run effects estimated a as i and i , with representing the convergence parameter (speed of adjustment to the long run relationship. Additionally, the short run coefficients are and . Based on the literature, in the long run the coefficients on physical and human capital are supposed to be positive, while that of population growth should be negative. The estimation of each equation takes into account the revenue neutral tax changes. That is, the greater use of one tax measure will inevitable reduces the amount of taxes needed to be raised from other tax measures. To account for this, one tax measure enters the equation at a time. The growth equation is subjected to the implicit constraint that the summation of all tax measures (ratio) is equal to one. This assumption ensures the estimation of the different tax policies under the assumption of revenue neutrality (Arnold et al (2011). Each equation is estimated by taking the general to specific approach. In that, four lags are included for the short run as shown in equation [2], after which the model is pared down. All equations are estimated with the Eviews 10 This highlights the point that cointegration does exist between variables integrated of varied orders . 11 package, using the OLS method and incorporating the Newey-West test. It is important to note that the Newey-West test produces consistent coefficients in the presence of heteroskedasticity and unknown correlation. Once the equations are estimated, the coefficient on the convergence variable ( y t 1 ) is expected to be negative and significant. Thereafter, various diagnostics tests are conducted to check for normality, long run relationship and stationarity of the residual. The residuals are tested for stationarity to avoid the production of spurious results. Additionally, the existence of a long run relationship between the variables is carried out by the use of a standard F-statistics to ascertain if the coefficients on the lagged level variables are jointly significant.11 The F-statistics is then compared against two sets of asymptotic critical values calculated by Pesaran, Shin and Smith (2001), which takes into consideration the level of significance and the number of explanatory variables in the equation (see Figure A in Appendix). These two set of critical values provide a band for the variables whether they are I(0), I(1) or fractionally integrated. In this regard, the hypothesis tested is as follow: H0: The coefficients on the lagged level variables are not jointly significant H1: The coefficients on the lagged level variables are jointly significant If the calculated F-statistics falls outside the upper level of the band, the null hypothesis is rejected. However, if the F-statistics falls below the lower band of the critical value, the null hypothesis cannot be rejected. It is important to note that the test for a long run relationship is inconclusive if the test statistics falls within the band.12 5.0 Results and Discussions The result for the unit root tests is shown in Table IV in Appendix. The Granger causality test results indicate that a bilateral relationship exist between per capita GDP and direct taxation as 11 The Wald coefficient test is used to determine the F-Statistic, where the coefficients are set equal to zero. It should be noted that the asymptotic distribution of the F-Statisrtics is non-standard. 12 See Pesaran, Shin and Smith (2001) for further details. 12 well as between per capita GDP and international trade, G.C.T. on imports, S.C.T. and ‘Other Trade’ taxes (see Table V in Appendix). The other significant results from the Granger causality test points to GDP per capita influence the other variables, with the exception of the growth rate of the population which impacted GDP per capita. It is important to note that the bilateral relationship between GDP per capita and taxes from international trade is in line with the close relationship that exists between output growth and volume of international trade as indicated by Jones (2001). The results from the final estimates of the basic growth model with only physical & human capital and population growth (baseline model) as well as those including the tax indicators are included in Table VII in the Appendix. The findings capture short run and long run changes in GDP from changes in tax policies. The results from the robustness test for each model are displayed in the respective tables. In addition, based on the coefficient of variation, the models have the ability to explain between 52.0 per cent and 70.0 per cent of the variation in per capita GDP. 5.1 Growth Components: Baseline Equation From the baseline equation [1] the coefficient on the lagged per capita income is -0.115 (see Table VI in the Appendix). This indicates that 11.5 per cent of deviations from long run equilibrium level are corrected for in each quarter. As a result, it will take approximately nine quarters to ensure full correction. In the other estimates, the average correction time is nine quarters. Additionally, the presence of a long run relationship between GDP per capita and the determinants of growth is confirmed based on the F-statistics of 13.02, which falls above the upper band of the critical value of 4.01, thus signalling the presence of a long run relationship between the variables at the 5.0 per cent level of significance. The signs of the coefficients on the key indicators of growth as depicted in the baseline model are in line with expectation where the coefficients on physical and human capital are positive, while that on population growth is negative. It is important to note that the coefficients on human capital and population growth display significant impact on GDP per capita in the long run, 13 while physical capital did not exhibit a significant long run impact on per capita GDP.13 The long run value on human capital from the baseline equation [1] suggests that a per cent increase in that variable human capital will result in a 1.9 per cent increase in economic growth over time. Additionally, a per cent increase in the growth of the population will result in a 12.5 per cent contraction in per capita GDP in the long run. Although Arnold et al (2011) results indicate a negative impact on per capita GDP for the OECD, the magnitude was much smaller. In the short run, all the variables display significant results. In that, a one per cent increase in the population growth will result in a 1.2 per cent expansion in economic growth. While physical and human capital have negative impacts on per capita GDP, with a 10.0 per cent increase in these variables resulting in a 0.3 per cent and 5.7 per cent decline in per capita GDP, respectively. Based on the results for physical capital, one could opine that there has not been adequate investment in capital or the reinvestment of profits in Jamaica. This is in a context where many of the large businesses have repatriated their profits to the parent company or have invested heavily in Government bonds. Notably, 52.3 per cent of the variation in per capita GDP in the baseline equation is explained by the model. 5.2 Growth Model with Tax Measures The results from the growth model which includes the long and short run impact of increases in the different shares of tax revenue are shown in Tables VI, VII and VIII, in the Appendix. It is important to note that the various tax measures have different impacts on per capita GDP. 5.2.1 Direct and Indirect Taxation It is found that by increasing the share of revenue via indirect taxes the policy measure has a positive and significant impact on per capita GDP in the long run. In that, a 10.0 per cent increase in indirect taxes is expected to increase per capita GDP by 3.0 per cent and hence contribute to an expansion in the GDP per capita over time. The positive impact was in line with the findings of Arnold et al (2011) and Kneller et al (1999). The results from the growth model for Direct taxes (equation 2.1) show that increasing direct taxation had a negative impact on per 13 This is evident in all equations that include a tax variable. 14 capita GDP in the long run, however, this impact is not significant. The impact in the short run for both Direct and Indirect taxation are insignificant. 5.2.2 Consumption In the short run, the positive coefficient on the changes in the share of Consumption taxes indicates that increases in these taxes have a significant contributory impact on GDP per capita. In this regard, the result suggests that a 10.0 per cent increase in the share of these taxes, will lead to a 0.6 per cent increase in GDP per capita. It is seen that the positive effect of consumption taxes on GDP per capita is only mirrored in the SCT in the short run. The result shows that a 10.0 per cent increase in the share of SCT will lead to a 0.3 per cent increase in GDP per capita in the short run. The Consumption coefficient had a negative impact on GDP per capita, which was, however, insignificant in the long run (equation [3] Table VI in Appendix). An examination of the components of Consumption, GCT and SCT provided mixed and insignificant long run results ([4] Table VI in Appendix). 5.2.3 International Trade Based on equation [5], an increase in the share of International Trade taxes has a positive but not significant impact in the long run on per capita GDP (see Table VII in the Appendix). The components of Trade, that is, GCT collected from imports and ‘other trade taxes’, exhibited significant long run positive effects, while SCT on import showed no significant impact [equation 6]. A 10.0 per cent increase in the share of GCT on imports and other trade taxes will result in increases of 1.5 per cent and 1.1 per cent, respectively in GDP per capita. Within the short run, an increase in the share of taxes from International Trade has a significant positive impact on economic growth, with a 10.0 per cent increase resulting in a 0.4 per cent rise in GDP per capita. However, the components of International Trade taxes, in particular SCT on imports and ‘other trade taxes’ displayed significant negative impacts in the short run, while the positive coefficient on GCT is insignificant. In this regard, a 10.0 per cent increase in SCT on imports and ‘other trade taxes’ will lead to a 0.06 per cent and 0.75 per cent decline, respectively, in GDP per capita in the short run. 5.2.4 Income & profits 15 Within Income & Profits, the significant negative impact on GDP per capita is evident in P.A.Y.E. and the ‘interest’ category in the long run, with a 10.0 per cent increase in the share of these taxes resulting in respective contractions in economic growth of 2.5 per cent and 0.8 per cent in the long run [equation 8], (in Table VII in the Appendix). However, an increase in the share of taxes garnered from companies has a positive but insignificant impact on growth over time. Within the short run, significant impacts were only evident for P.A.Y.E. and ‘interest’, with each having a negative and positive impact, respectively. 5.2.5 Production & Consumption Based on the results, an increase in the share of taxes collected from Production & Consumption [equation 9] has a positive, however, insignificant impact on per capita GDP in the long run (see Table VIII in Appendix). Similar results are found for the short run. Within the Production & Consumption group, only the ‘licensees’ category exhibited a significant positive impact on per capita GDP in the long run, but a negative and significant impact in the short run. In this regard, a 10.0 per cent increase in the share would lead to a 6.2 per cent increase and 0.6 per cent decline in GDP per capita, in the long run and short run, respectively. 5.3 Growth Model with Taxation Index This section presents the findings derived from the taxation indexes, which assess the impact of the percentage change(s) in the tax rate on changes in per capita GDP. The P.A.Y.E. index produced significant results in both the long run and short run. In the long run a 10.0 per cent increase in the share of revenue from P.A.Y.E. will account for approximately a third of the increase in per capita GDP. Within the short run, a 10.0 per cent increase will result in a 1.5 per cent decline in per capita GDP. The GCT index had a positive but insignificant impact in the long run. However, the impact in the short run is positive and significant, with a 10.0 per cent increase expected to increase per capita GDP by 0.5 per cent. In comparing the results from the indexes and tax measures as a share of tax revenue models, P.A.Y.E. in both cases have similar magnitude and significance in the long run but in opposite direction. In that, within the indexes model, increases in P.A.Y.E. have a positive impact on per capita income, which is not in line with theory. For the short run, both models suggest negative 16 and significant results. Within the short run, the GCT index was positive and significant, however, it was not significant in the tax revenue measure model. 5.4 Discussion All the significant results from the impact of the tax measures on GDP per capita are summarize in Table 2. These results facilitate the creation of a tax and growth rank, that is, tax groups that have the least to most harmful impact on GDP per capita. The results show that indirect taxation has a positive and significant impact on economic growth in the long run. In this regard, tax policies geared towards indirect taxes would be beneficial, not only to increasing Government revenue, but increasing GDP per capita thereby stimulating long run economic growth. This result is collaborated with the findings of Kneller et al (1999). In the long run, based on the tax growth rank, revenue garnered from licences have the largest positive and significant impact on per capita GDP, followed by GCT on imports then ‘other international taxes’ (see Table 2). Interest on earnings has the smallest negative impact on per capita GDP in the long run followed by P.A.Y.E. The results suggest that an increase in the share of taxes from P.A.Y.E. has the greatest harm to economic growth in the long run and any distortion to GDP per capita will take approximately nine (9) months to be corrected. Furthermore, for a country such as Jamaica that is in its recovery phase of its growth cycle, it would be unwise for policymakers to attempt to increase revenue intake from this measure. Arnold et al (2011) suggests that it would be more beneficial to reduce the taxes on personal income with the aim of increasing spending and further stimulate economic growth. In this context, within the short run (as shown in Table 2), consumption tax has the greatest beneficial impact on growth, in particular from SCT. It is important to note that the GCT index does indicate strong impact in the short run. Interest on earnings essentially has no impact on per capita GDP in the short run. PAYE, licences and local production taxes as well as ‘other international trade’ taxes have negative impacts in the short run. Notably, International Trade taxes have an overall positive impact on growth in the short run. 17 Table 1: Significant Results from Equations with Tax Measures Variables Indirect Consumption SCT International Trade GCT on Imports PAYE Other international Taxes Interest Licences PAYE-Index GCT-Index Impact Long run Short run positive* --------positive*** ----positive*** ----positive** positive*** ----negative *** negative* positive** negative*** negative*** positive* positive*** Negative* positive* ----- negative*** positive*** * Significant at 10.0 % level; ** at 5.0 %; *** at 1.0 % level. See Tables VI, VII and VIII for details. 6.0 Conclusion & Recommendations The paper set out to examine the impact taxation has on economic growth, using quarterly data in an autoregressive distributed lag framework. The results indicate that by increasing the share of tax revenue garnered from indirect taxes, in particular GCT on imports, policymakers actions would result in long run economic growth benefits. However, if policymakers are myopic or interested in the impact on economic growth in the short run, increasing the share of tax revenue from consumption taxes, in particular SCT would be conducive to growth. The results show that any policy action aimed at increasing the P.A.Y.E. tax would have a negative and significant impact on GDP per capita over time. Additionally, this negative impact will take approximately 9 quarters to correct. Greater benefit would be derived by reducing the taxation on personal income if the objective is to stimulate demand. With the country currently in a recovery phase, tax policy strategy aimed at stimulating demand is important. As such, any strategy should be guided by the findings of this study. The monetary authority will play a vital role in maintaining price stability in light of the expected pickup in demand. Additionally, the potential for growth in the Jamaican economy, at a time when the 18 Central Bank policy rate is at its lowest, 6.25 per cent, is an added factor that will aid in improving the fiscal position. This is in a context that the differential between the rate of growth and interest rate has narrowed, which is a vital ingredient for the Government to achieve its goal of fiscal and debt sustainability. 7.0 References Arnold, Jens, Bert Brys, Christopher Heady, Asa Johansson, Cyrille Schwellnus and Laura Vartia. “Tax Policy for Economic Recovery and Growth,” The Economic Journal 121 (2011): F59–F80. Accessed April 01, 2011. Doi: 10.1111/j.1468-0297.2010.02415.x. Benhabib, Jess and Mark Spiegel. 1994. The Role of Human Capital in Economic Development Evidence from Aggregate Cross-country Data. Journal of Monetary Economics 34 (1994) 143173. Accessed April 01, 2011. http://www.sciencedirect.com/science/article/pii/0304393294900477 Charles Jones, Introduction to Economic Growth (New York: W.W. Norton &Company Inc., 2002), 237. Földvári1, Péter and Bas van Leeuwen. 2009. An Alternative Interpretation of Average Years of Education in Growth Regressions. http://www.iisg.nl/indonesianeconomy/humancapital/workingpapers/av_years.pdf accessed: April 27, 2011 Greenidge, Kevin and Lisa Drakes. 2009. Tax Policy and Macroeconomic Activity in Barbados. http://www.centralbank.org.bb . Accessed April 10, 2011. Kneller, Richard, Michael, Bleaney and Norman Gemmell. 1999. Fiscal Policy and growth: Evidence from OECD Countries. Journal of Public Economics 74 (1999) 171-190. Ministry of Finance and Public Service Jamaica. www.mof.jm.org. Pesaran, M. Hashem and Yongcheol Shin. 1997. An Autoregressive Distributed Lag Modelling Approach to Cointegration Analysis. http://www.econ.cam.ac.uk/faculty/pesaran/ardl.pdf. Accessed April 2011. Pesaran, M., Yongcheol Shin and Richard Smith. 2001. Bound testing Approaches to the Analysis of Level Relationships. Journal of Applied Econometrics 16: 289–326 (2001) DOI: 10.1002/jae.616. http://onlinelibrary.wiley.com/doi/10.1002/jae.616/pdf. Accessed April 01, 2011. 19 Romer, Christina and David, Romer. 2010. The Macroeconomic Effects of Tax Changes: Estimates based on a New measure of Fiscal Shocks. ] http://elsa.berkeley.edu/~dromer/papers/RomerandRomerAERJune2010.pdf . Accessed April 01, 2011. Statistical Institute Jamaica. http://statinja.gov.jm/ 8.0 Appendix Table I: Variables and Measurement Variables GDP per capita Measure/Proxy Data Sources GDP per head of working age population 14 years and over Statistical Institute of Jamaica (STATIN) Physical Capital Gross capital formation share of GDP STATIN Stock of Human Capital Quarterly primary institution enrolment, calculated by interpolating annual primary institution enrolment14 Economic and Social Survey of Jamaica (ESSJ) Population growth Growth rate of the working age population ESSJ Tax Indicators Percentage Share of Tax Revenue: Ministry of Finance & Public Service, Jamaica. Determinants of Growth: (X) Income & Profit: P.A.Y.E., bauxite Companies and Interest (Tax on dividend) & Other Production & Consumption: GCT (local), SCT (local) and licenses Trade: GCT (import), SCT (import) and Other International trade Consumption: GCT and SCT Direct and Indirect Taxation Tax Indexation: P.A.Y.E. and GCT 14 Using quadratic matched-average method in EVIEWS 20 Table II: Definition of Variables Name y K H wkp Inc paye comp int prdcon gctlcl sctlcl lice Definition Gross Domestic Product per capita Physical Capital Accumulation Stock of Human Capital Population Growth Tax Indicators Income & Profit Pay as you earn (P.A.Y.E.) Bauxite and Other Companies Tax on Dividend, other individuals and tax on interest Production and Consumption General Consumption Tax (local) Special consumption tax (local) Motor vehicle licenses, other licenses, ‘betting, gaming & lottery’, education tax, contractors levy and stamp duty (local) trade gctimp sctimp othint consp gct sct direct indtax International Trade General consumption tax (imports) Special consumption tax (imports) Custom duty, stamp duty and travel tax Consumption General consumption taxes Special consumption taxes Direct taxes Indirect taxes indexpaye indexgct Indexation of P.A.Y.E. Indexation of General consumption tax 21 Table III: Descriptive Statistics Standard Error Variables Mean GDP per capita Primary Institution Enrollment Working Population Capital Accumulation/GDP -0.22 Direct Taxes Indirect Taxes 47.07 52.93 Consumption 34.99 0.64 15.42 48.42 International Trade Income & Profit Production & Consumption 27.88 40.60 30.91 0.43 0.61 0.26 18.65 30.46 25.22 42.22 72.74 36.76 -0.13 0.30 -0.86 Minimum Maximum Growth (%) 2.65 -7.30 5.88 0.89 0.50 17.14 -3.09 -0.54 -41.31 Share of Tax Revenue (%) 0.60 36.71 0.60 20.91 3.29 3.25 41.20 79.09 63.29 Source: Author’s calculations 22 Table IV: Correlation and Order of Integration of Variables Correlation of variables with GDP per capita GDP per capita Population Growth Physical Capital Human Capital Stock Trade GCT Imports SCT Imports Other International Consumption GCT SCT Production & Consumption GCT (local) SCT (local) Motor vehicle licenses, other licenses, ‘betting, gaming & lottery’, education tax, contractors levy and stamp duty (local) Income & profit P.A.Y.E. Bauxite and Other Companies Tax on Dividend, other individuals and tax on interest Direct Taxation Indirect Taxation GCT Indexation PAYE Indexation Source: Author’s calculations Order of Integration 1 -0.072 0.073 0.001 I(1) I(0) I(0) I(1) -0.1783 -0.184 -0.074 -0.217 -0.255 -0.146 -0.184 I(0) I(0) I(0) I(0) I(1) I(0) I(1) I(0) -0.323 -0.013 -0.259 -0.306 0.261 0.047 I(0) I(1) I(0) I(0) I(0) I(0) 0.241 I(0) 0.122 0.243 -0.200 -0.072 I(0) I(0) I(0) I(0) 23 Table V: Results from the Granger Causality Tests Pairwise Granger Causality Tests Sample: 1990Q1 2010Q4 Lags: 2 Null Hypothesis: Obs F-Statistic Prob. DLINC does not Granger Cause DLY DLY does not Granger Cause DLINC 81 2.13376 3.98448 0.1254 0.0226 DLCOMP does not Granger Cause DLY DLY does not Granger Cause DLCOMP 81 1.07329 6.44106 0.347 0.0026 DLINT does not Granger Cause DLY DLY does not Granger Cause DLINT 81 4.85851 1.16227 0.0103 0.3183 DLPRDCON does not Granger Cause DLY DLY does not Granger Cause DLPRDCON 81 0.24296 4.12554 0.7849 0.0199 DLLICE does not Granger Cause DLY DLY does not Granger Cause DLLICE 81 1.73568 4.85481 0.1832 0.0104 DLTRADE does not Granger Cause DLY DLY does not Granger Cause DLTRADE 81 8.06038 10.4589 0.0007 0.0001 DLGCTIMP does not Granger Cause DLY DLY does not Granger Cause DLGCTIMP 81 6.75867 5.90921 0.002 0.0041 DLSCTIMP does not Granger Cause DLY DLY does not Granger Cause DLSCTIMP 61 3.09886 1.09506 0.0529 0.3416 DLOTHINT does not Granger Cause DLY DLY does not Granger Cause DLOTHINT 81 8.72831 15.5679 0.0004 0.00 DLCONSP does not Granger Cause DLY DLY does not Granger Cause DLCONSP 81 1.92338 2.71158 0.1532 0.0729 DLSCT does not Granger Cause DLY DLY does not Granger Cause DLSCT 81 2.57251 5.8611 0.083 0.0043 DLGCT does not Granger Cause DLY DLY does not Granger Cause DLGCT 81 2.37053 2.69856 0.1003 0.0738 DWKP does not Granger Cause DLY DLY does not Granger Cause DWKP 81 4.7478 1.7568 0.0114 0.1795 DLDIRECT does not Granger Cause DLY DLY does not Granger Cause DLDIRECT 81 2.91872 4.20681 0.0601 0.0185 DINDEXGCT does not Granger Cause DLY DLY does not Granger Cause DINDEXGCT Calculated with EViews 81 4.60917 1.75888 0.0129 0.1792 24 Table VI: Results for taxation and Growth Dependent Variable: GDP per capita Variables Baseline (1) Long run Impact C Baseline Model LKt-1 LHt-1 WKPt-1 0.026 1.939*** -12.452*** Tax Structure Variables LDIRECTt-1 LINDIRECTt-1 LCONSPt-1 LGCTt-1 LSCTt-1 Direct Taxation [2.1] Indirect Taxation[2.2] 0.003 1.884*** -15.532**** 0.001 1.953*** -16.385*** Consumption [3] 0.055 1.936*** -15.809*** Components of Consumption [4] -0.061 4.576*** -24.758*** -0.321 0.238* -0.036 0.061 -0.015 Short run Impact DLYt-2 DLKt-3 DLKt-4 DLHt-2 DWKPt-1 DLDIRECTt DLINDIRECTt DCONSPt-1 DLGCTt-1 DLSCTt-1 Adjusted R2 S.E. of regression Durbin-Watson Wald test: F-Statistics Breusch-Pagan ARCH Residual Stationary Norm (prob) -0.621*** -0.034** -0.568* 1.175*** -0.625*** -0.646*** -0.638*** 0.038*** 0.039*** -0.388* 1.407*** 0.43*** -0.443* 1.628*** -0.383* 1.405*** -0.594*** -0.053*** -0.716** 1.261*** 0.006 -0.004 0.058*** -0.01 0.033*** 0.523 0.018 2.151 8.968 [0.0056] 1.523 [0.2047] Yes 2.5 [0.287] 0.561 0.0178 2.058 7.097 [0.0588] 0.559 0.018 2.094 7.824 [0.0534] 2.031 [0.0994] Yes 1.801 [0.1384] Yes 9.184 [0.010] 10.073 [0.0065] 0.571 0.018 2.103 7.25 No [0.0398] 0.5981 0.582 0.017 2.068 5.702 [0.6194] 0.8610 Yes 14.69 [0.000] Yes 8.067 [0.0135] Notes: L- denotes the log of the variable. D – denotes the first difference of the variables * Significant at 10.0 % level; ** at 5.0 %; *** at 1.0 % level. Norm: normality test of the residual based on probability value from the Jarque-Bera test statistics. The null hypothesis of a normal distribution. Wald test coefficient test for long run relationship. Heteroskedasticity Test: ARCH. The null hypothesis test suggests no Heteroskedasticity. 25 Table VII: Results for taxation and Growth Variables Long run Impact Baseline Model LKt-1 LHt-1 WKPt-1 Tax Structure Variables LTradet-1 LGCTIMPt-1 LSCTIMPt-1 LOTHINTt-1 LINCt-1 LPAYEt-1 LCOMPt-1 LINTt-1 Short run Impact DLYt-1 DLYt-2 DLKt DLKt-3 DLKt-4 DLHt-2 DLHt-3 DWKPt DWKPt-1 DLTRADEt-1 DLGCTIMPt-3 DLSCTIMPt-2 DLOTHINTt-3 DLINCt-1 DLPAYEt-2 LCOMPt DLINTt-4 Adjusted R2 S.E. of regression Durbin-Watson Wald test: F-Statistics Heteroskedasticity Residual Stationary Norm (prob) International Trade (5) International Trade Components (6) Income & Profit (7) Income & Profit Components (8) 0.044 1.956*** -10.184*** -0.055 0.485*** 0.093 -0.009 1.875*** -11.696*** -0.027 1.823*** -8.667*** 0.053 0.152*** 0.000 0.114** -0.241 -0.253*** 0.022 -0.081*** 0.346*** -0.534*** -0.585*** -0.599*** -0.036*** -0.031*** -0.584** -0.693*** 1.236*** 1.194*** -0.021 -0.028** -0.537* -1.028*** 1.445*** 0.041** 0.877** 0.029 -0.006* -0.075*** 0.008 0.545 0.018 2.21 6.26 [0.0735] Yes 4.204 [0.122] 0.696 0.014 1.697 31 [0.8177] Yes 0.765 [0.682] 0.533 0.018 2.064 7.162 [0.021] Yes 2.812 [0.245] -0.034* 0.004 0.007* 0.548 0.018 2.184 6.704 [0.0166] Yes 1.064 [0.588] 26 Table VIII: Results for taxation and Growth Variables Long run Impact LKt-1 LHt-1 WKPt-1 LPRDCONt-1 LGCTLCLt-1 LSCTLCLt-1 LLICEt-1 LINDEXPAYEt-1 LINDEXGCTt-1 Short run Impact DLYt-2 DLKt-3 DLKt-4 DLHt-2 DLHt-3 DWKPt DWKPt-1 DLPRODCONt-4 DLGCTLCLt-3 DLSCTLCLt-4 DLLICEt-1 DINDEXPAYEt-2 DINDEXGCTt-2 Adjusted R2 S.E. of regression Durbin-Watson Wald test: F-Statistics Heteroskedasticity Residual Stationary Norm (prob) Production & Consumption (9) Production & Consumption Components (10) -0.032 2.021*** -21.723*** 0.489 -0.131 1.071 -13.970*** PAYE Index (11) 0.009 1.882*** -16.418*** GCT Index (12) 0.017 1.925*** -10.817*** -0.162 0.091 0.616** 0.300* 0.008 -0.598*** -0.029** -0.652*** -0.661*** 0.03* 0.032*** -0.382 -0.503** -0.616** 1.248*** -0.047 1.184*** -0.648*** -0.021 -0.705** 1.306*** 1.44*** -0.006 -0.007 -0.058* -0.154*** 0.054*** 0.548 0.018 2.185 13.024 [0.0153] Yes 11.26[0.004] 0.604 0.017 1.856 7.79 [0.0789] Yes 22.02[0.000] 0.57 0.018 2.016 7.42 [0.0278] Yes 10.16[0.006] 0.552 0.018 1.983 10.63 [0.058] Yes 9.64 [0.008] 27 Figure A: F-Statistics for Long run Relationship Source: Pesaran et al (2001) Bounds Testing Approaches to the Analysis of Level Relationships, page 300. K- number of explanatory variables 28